AI is making your best people faster. And your worst processes worse.

1 in 10 organisations have actually scaled AI. The other nine bolted it on.

Almost half of executives say AI has done nothing for their bottom line. I believe them. And I don't think it's AI's fault.

I picked up the valuable stats at the London tech Week last week. Accenture's Generating Impact report (April 2026) has the number: 46% say AI has delivered little or no impact on their P&L. Set that next to another finding in the same report and the picture snaps into focus. Only 1 in 10 organisations has actually scaled AI into how the business runs.

So this isn't a story about AI underdelivering. It's a story about where the gains get stuck.

The gains are real. They're just personal.

Let's be fair to the technology. It works. Nearly one in five UK workers now uses gen AI every day, triple the figure two years ago. Accenture reckons 82% of working hours could be enhanced by it, up from 47%. People are genuinely faster.

I'm Exhibit A. My own productivity has gone somewhere I couldn't have imagined two years ago. I've got a stack of AI agents working across my day now, wired into my tools and data through MCP. On a personal level, it's changed what one person can get done.

But that's the trap. All of that is me. None of it, on its own, changes how my business creates value. And a P&L doesn't measure how fast individuals work. It measures how the organisation creates value.

That's the gap. And it's the whole story.

The leap is encoding, not buying

Here's where I see it most clearly: my own sales team.

We're at a point now where every one of us is using AI in genuinely clever ways. Individually, we're flying. But the next step, the one that actually shows up in the numbers, isn't more tools. It's bringing together the best of what each person is doing and encoding it as a team.

Because right now there's drift. The same service gets described in subtly different language, priced in slightly different ways, presented in different branding, depending on who's at the keyboard. Brilliant individual work, no shared system underneath it. The value is real, it's everywhere, and it's trapped at the level of the person.

Getting it out means doing the unglamorous work: re-engineering the process, agreeing the one right way, and encoding it into shared workflows. Then the harder part, the people part. The training and the adoption that turn a clever individual habit into how the whole team operates.

That half doesn't demo well. So it rarely gets funded. Everyone buys the tool and skips the encoding, then wonders why the P&L looks the same.

That skipped half is the entire game.

It's not just us

In my work with Siloy, I see the same pattern across customers right the way through Europe. Pockets of brilliant early adopters, experimenting, getting real personal wins. Experimenting is good. It's where it has to start.

But almost nobody has done the next bit: the hard yards of operationalising it, encoding it, making it the way the organisation works rather than the way a few clever people work. Everyone's stuck at the same place on the same curve. The experiment is easy. The encoding is hard.

Bolt it on and you don't get nothing. You get worse.

And there's a sting for anyone tempted to skip the redesign and jump straight to deployment. Drop AI onto a broken process and the process doesn't sit still. It runs faster. It produces more of the wrong output, at scale, more confidently. And it bakes the bad process in, because now it's encoded and harder to pull apart.

A bad process with AI on top isn't a bad process plus a bit of upside. It's a bad process with the volume turned up.

The lever moved

Matt Prebble at Accenture put it more cleanly than I can: "The productivity lever has shifted from the technology itself to leadership and how organisations are designed to use it."

That's it exactly. The bottleneck was never the model. The tools are finally ready, genuinely ready, in a way they weren't two years ago. The bottleneck is whether we'll do the organisational work: reimagine the process, encode the best of what people are already doing, bring everyone with us, and deploy it across the system rather than handing it out person by person and calling that a strategy.

Personal productivity is where AI starts. 
It is not where the value is.

The real prize comes from encoding it into how the whole organisation works. That's slower, harder, and a lot less exciting than buying another tool. It's also the only thing that ever moved the number.

We've started doing that hard work on our own team. I know which work I'd rather be doing.